When I upgrade Oracle GI clusterware and I monitor the log file of the rootupgrade.sh script that my sysadmin is running, I see something interesting. Something called "crsctl status sandbox". I realize it's internal but what is sandbox? Is the upgrade running a "dry run" in memory or something? Just curious. Also, what is the point of no return? I know that rootupgrade.sh is restartable and I know that script actually updates the OCR and the cluster components, but what step, I believe there are 18 total in the 19c upgrade, after what step if it fails can I still restart the old previous version clusterware and return the cluster to a stable running state until I can let's say raise an SR and attempt to upgrade again at a later date?